This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
8/11/2019 How to Excel at Research Computing in Times of Diminishing Resources, Growing Demand, and Expanding Possib…
investments contribute both directly and indirectly to economic growth, although estimates vary
widely and depend on the methodology used.4
When considering academic research today, three noteworthy facts surface:
1. Basic research has a 25-year payoff, from basic science findings to practical
applications. Our current economic climate is not conducive to funding activities thatmight or might not deliver benefits 25 years hence.
2. Research today is highly dependent on technology.
3. The United States’ leadership in research is waning. That is not solely due to the Great
Recession. Even in 2007, a National Academy of Sciences report warned, “Having
reviewed trends in the United States and abroad, the committee is deeply concerned
that the scientific and technological building blocks critical to our economic
leadership are eroding at a time when many other nations are gathering strength.”5
The special collaboration between government and higher education that was largely
responsible for the United States’ current leadership in academic research is unraveling at a
time when other countries are building research and developmentinfrastructures. Corporations are not stepping in to take up the slack.
Growth in academic R&D continues today, but its pace has slowed,
from annual increases averaging 14% to 19% in the 1950s and ’60s
to about 10% in the 1970s and ’80s to only 7% in the first decade of
this century.6 Major challenges facing academic research in the
United States include:
Weakening partnerships among government, businesses, and higher education
Declining federal and state funding
Rising accountability demands for higher education institutions
Declining opportunities for new faculty
Underinvestment in campus infrastructure, particularly in cyberinfrastructure
Increasing regulatory and reporting requirements
Sponsored research that funds only part of research costs, leaving a gap that institutions
need to cover
Increasing international competitiveness—in research and for students7
Academic research is today experiencing decreasing resources, investment, and leadership;
increasing costs and regulatory and accountability requirements; and expanding requirements
and needs for technology. It was against this backdrop that ECAR Annual Meeting attendeesconsidered the opportunities and challenges facing higher education IT leaders. These
proceedings distill the contributions of the meeting’s speakers as well as attendee comments.
(See the appendix for a complete listing of sessions and speakers.)
Basic research has a
25-year payoff, from
basic science findings to
practical applications.
8/11/2019 How to Excel at Research Computing in Times of Diminishing Resources, Growing Demand, and Expanding Possib…
Increases in computational capability can fundamentally change even the problems faculty
want to attack. New technologies are changing scholarship and research for all faculty, not just
those in traditional scientific computation areas or those focused on technology. Faculty have a
clear notion of the new modes of research that they must pursue to be competitive. As they
embrace research computing, they are starting to challenge central IT to move faster and to
understand e-science more deeply.
Specialized computers and interesting technologies—like Google Glass—abound. Some of
them will be critical to research computing. For example, quantum computers are in the
exploratory phase now, and they are extremely effective for certain types of database queries.
Using Technology to Engage Students in Research
Research computing can extend powerfully into instruction. Researchers at USC created an
internship program for undergraduates to contribute discipline-specific software to high-
performance computers.8 The students use the same equipment as the faculty to meld
research experience and classroom activity; they are part of the research program. Integratingresearch computing into courses (with sufficient funding) can elevate awareness of research
computing resources and the science they support. Future steps could extend to granting
certificates or minors in the use of emerging research technology in such areas as data
analytics, high-performance computing (HPC), quantum computing, and visualization. Students
with research computing credentials will likely be very attractive to employers, who are
increasingly investing in such capabilities.
Middlebury College’s Digital Liberal Arts program gives students an opportunity to meaningfully
engage in research. It creates resources for use within curriculum, fosters media literacy, and
gives student participants opportunities to engage in issues such as preservation, use of
technology for analysis, and representation of analysis. Michael Roy, dean of library and
information services and CIO of Middlebury College, noted that
a number of digital liberal arts research projects were driven by the desire to create resources
in the teaching process and were then expanded and used by others. The faculty who engage
students in this sort of work report that developing these capabilities through these sorts of
projects is important for supporting the liberal arts mission and educating students to be
lifelong learners.
Expanding the Possibilities in the Humanities
The humanities have been embracing computation as an emerging form of scholarship,
particularly over the past 10 years. Examples from Middlebury College faculty include:
A digitized, geo-coated map of the Gettysburg battlefield according to Robert E. Lee’s
perspective to provide further analysis of the battle9
A WordPress site created as a platform for peer-to-peer reviews of book chapters to
augment the traditional editor-mediated peer-review process10
The Mahri Poetry Archive, a structure to collect, digitize, and share audiotapes of Yemeni
people reciting Mahri poetry, an exclusively oral form of poetry11
8/11/2019 How to Excel at Research Computing in Times of Diminishing Resources, Growing Demand, and Expanding Possib…
Research computing services need to support both traditional quantitative research and
qualitative scholarship. Michael Roy suggests that “digital humanities looks a lot like research
computing in the vast majority of methods.” The tools to support those methods are fairly
familiar to anyone who is working in the computing and library or curricular computing
environments. Four areas distinguish digital humanities
from other types of research computing support: toolbuilding, curation and preservation, new modes of
publication, and evaluation structures. Scientific research
tends to be collaborative and team-based; humanities
research tends to involve a singular, individual scholar,
perhaps with graduate students.12
Crowdsourcing and Collaborations
With the aid of technology, distances are no longer a barrier to collaboration. Science today is
no longer a single-investigator activity, but instead involves multiple collaborators often working
in different disciplines, different institutions, and different countries. IT can facilitate this kind of
activity in numerous ways, whether supporting distance collaborations with videoconferencingand hosting services, providing high-speed access to external resources, delivering local
services globally, or supporting collaboration applications. Tools such as the Electronic
Laboratory Network (ELN) are growing in popularity. ELNs are wiki-like document or protocol
repositories with better tags, date/time stamp, and digital tools for time and version control that
can promote collaboration. They can be hosted on campus or in the cloud.
Collaborations come with IT-related challenges. They include the obvious bandwidth and
access challenges, as well as security and even regulatory challenges. One non-U.S.
participant described faculty concerns about collaboratively passing data across the Internet,
especially when it crosses international borders. IT departments must understand researchers’
needs and the tools they are already using. Individual researchers can now afford some ELN
solutions, but their data-protection capabilities may not meet institutional standards.
Big Data Placing Big Demands on IT
New avenues of research and scholarship have opened up as a result of technology ’s ability to
collect, store, process, and depict massive amounts of structured and unstructured data.
Scientists and scholars in fields from astronomy to medicine to history are able to answer
previously unanswerable questions. Technology has cracked the big-data nut, and that is
posing challenges for IT.
Networking
Advances in data storage, retrieval, and processing have enabled researchers to work with
huge volumes and varieties of data. The volume of big data is increasing by orders of
magnitude. As research becomes increasingly data intensive, it demands higher throughput
and faster access from networks. Examples of solutions13 include:
Software-defined networking, which can simplify network management and help address
challenges involved with big data and security. NYSERNet’s work to implement
OpenFlow across the NYSERNet network is an example.
“ Digital humanities looks a lot likeresearch computing in the vast
majority of methods.”
—Michael D. Roy, Middlebury College
8/11/2019 How to Excel at Research Computing in Times of Diminishing Resources, Growing Demand, and Expanding Possib…
The Science DMZ.14 Institutions used to attach the research network to the administrative
or campus network. This model is reversing, with the research network as the primary
Internet-facing network and the administrative network configured almost like a small
local area network. The Science DMZ involves connecting supercomputer centers or
large HPC clusters directly to the Internet using data transfer node (DTN) devices that sit
outside the firewall and are designed to transmit data very quickly. The DMZ modelallows institutions to optimize network equipment, configuration, and security policies for
research computing.
Solutions like Globus,15 a file server with carefully crafted network interfaces for high-
speed file transfer in the cloud. One potential best practice discussed at a recent
conference is for each research university to have a Globus online host to move large,
multiterabyte data sets across the network from campus to campus.
Regional optical networks like NYSERNet, IRON, and KyRON are a major part of the higher
education networking topology. They are currently experiencing significant funding challenges,
creating the potential for bottlenecks between a national Internet2 100-gigabit backbone
network structure and increasingly effective Science DMZs.
Data Management
As noted by Caruso et. al, big data is also generating data-management challenges including
metadata for searching, standards, preservation, archiving, reproducibility, and retrieval.
Institutions need to develop practices and policies for data preservation, security,
standardization, and efficient storage.
As researchers create more data, it is increasingly imperative to create intentional strategies
around which data to keep or throw away—and to do this up front as part of a research project
life-cycle planning phase, perhaps adopting phases familiar to researchers, like preaward,
active award, and postaward. Some institutions are creating templated data-management
language in grants. IT is engaging libraries in the process because they understand the data,the disciplines, and data curation, as well as sustainability. Sustainability issues include
standards and long-term preservation. Libraries have experience with platforms, formats, and
cataloging, all components of data archiving and management. Long-term preservation is a
particular challenge because there is no clear funding model. One thought is to consider
endowing data sets.
Creating an institutional one-size-fits-all plan may not be a good approach. The data challenges
are different for different domains. As a result, discipline groups have begun to address these
issues for their specific domain. Perhaps partly for this reason, there are no cohesive data-
management requirements among federal agencies, creating another impediment to a single
institution-wide solution.
16
The Council on Library and Information Resources (CLIR) has created an initiative called the
Committee on Coherence at Scale to foster strategic thinking about how to more rigorously
manage the transition from analog to digital in higher education.17 Among the ideas emerging in
this space is applying the NSF’s cyberinfrastructure model to the humanities and social
sciences, resulting in centers resembling supercomputer centers—pockets of expertise and
coordinated activities.
8/11/2019 How to Excel at Research Computing in Times of Diminishing Resources, Growing Demand, and Expanding Possib…
The ubiquity of data and people who wish to easily access and manipulate it lead to
information-security challenges that IT needs to address. In the absence of strong security,
some researchers are even opting to store data on servers that are not connected to the
network, to keep them safe from intrusions.
Commercial cloud services bring their own security challenges. One solution is to buy a site
license that enables faculty to encrypt data and store it in the cloud. With this arrangement,
both faculty and central IT possess an encryption key to manage the university assets. NET+ is
looking at services to support HIPAA compliance, such as cyberclouds that encapsulate data.
Maturing E-Science Services to Advance Research:A Service Management Approach
College and university leaders are looking to IT leaders to do more with less —or at least with
the same. Research computing is no exception to that trend. Many of the solutions either in
place or being explored to balance resources, expectations, and requirements in other areas ofIT support may be equally appropriate for research computing. But they will need to be adapted
to the unique facets of this mission.
E-Science Support Today
Computing in support of the research mission accounts for less than 1% of most institutions’
central IT budgets but reaches 4% or higher in 25% of institutions; such support exceeds 12%
in 1 in 10 institutions. Even the institutions devoting the most IT support to the research
mission—public doctoral institutions—spend an average of only 11% of the central IT expenses
on research. In comparison, research accounts for an average of about 11% of institutional
expenses in all four-year institutions (12.1% for public universities and 10.8% for privates). 18 So
the average level of institutional spending on research—11% to 12%—is a level that only 10%of institutions achieve in central IT research computing spending.
Research is arguably the most highly decentralized of the three major missions of academic
institutions—teaching and learning, research, and administration. That might at least partly
explain why research computing seems to be underfunded compared with overall research
expenditures. In fact, about two-thirds (64%) of institutions that support research computing
have departments that independently provide research computing services.
Research computing appears even less coordinated when looking at individual services. On
average, central IT manages any individual research service in 25% of institutions offering that
service. Even with an expanded centralized role to include system oversight (1%) or shared
responsibility (32%), management of an individual research service is decentralized in anaverage of 29% of institutions. This is probably due largely to the sheer complexity of e-science
service delivery, which ranges from technically complex services such as HPC and grid
computing to knowledge-intensive consulting in visualization, grant preparation, statistics, and
access to federally funded research resources (see figure 1).
8/11/2019 How to Excel at Research Computing in Times of Diminishing Resources, Growing Demand, and Expanding Possib…
ECAR has developed a maturity index for research computing20 that enables institutions to
chart their progress along relevant dimensions of culture, infrastructure, investment, and central
IT involvement and service quality (see figure 3).
Figure 3. Research Computing Maturity Index
Business Models: Funding Is Critical to Success
Funding may be the biggest barrier to effective and efficient research computing services. Many
IT departments find themselves having a funding conversation around every instance of
research service provision. One solution is to align the IT
organization with institutional academic initiatives. If IT is
part of a campus initiative that puts some money on the
table, the dialogue with faculty is different and much
friendlier.21
The current university business model for provisioninghigh-performance computing, in which vendors sell
equipment directly to researchers and each researcher ’s
HPC facility is a “tub on its own bottom,” is not cost-
effective. An alternative business model at Purdue
University is to manage HPC as a centralized service, with
service level agreements designed to address faculty
needs. This new model requires that researchers can trust
central IT to build a competitive research computing
environment and meet their expectations. As Gerry McCartney, Purdue’s CIO, aptly put it, “The
academic culture is a coalition of the willing. Once you make something a rule, you lose a lot of
good will. Faculty are rational; they come for the prices and stay for the service. They will notcome for the service because they won ’t believe you can provide it. If you offer them a great
deal and then provide them with service, they will stay.”
Recommendations resulting from the success of the Purdue model include:
Use faculty references to gain faculty credibility. A senior faculty member ’s
endorsement of the buying program was more effective than any CIO-authored memo.
“ The academic culture is a
coalition of the willing. Once you
make something a rule, you lose
a lot of good will. Faculty are
rational; they come for the prices
and stay for the service. They
will not come for the service
because they won’ t believe you
can provide it. If you offer them a
great deal and then provide
them with service, they will stay.”
—Gerry McCartney, Purdue University
8/11/2019 How to Excel at Research Computing in Times of Diminishing Resources, Growing Demand, and Expanding Possib…
3. BankBoston Economics Dept., “MIT: The Impact of Innovation,” http://newsoffice.mit.edu/1997/study-0305.
4. Charles I. Jones, “Sources of U.S. Economic Growth in a World of Ideas,” American Economic Review 92, no. 1(March 2002), 220 –239; U.S. Congress Joint Economic Committee, The Pivotal Role of Government Investment
in Basic Research, May 2010; and National Research Council, “Foreign Participation in U.S. Research and
Development: Asset or Liability?” (Washington, DC: The National Academies Press, 1996),
http://www.nap.edu/catalog.php?record_id=4922.
5. National Research Council, “Rising Above the Gathering Storm: Energizing and Employing America for a Brighter
Economic Future” (Washington, DC: The National Academies Press, 2007),
http://www.nap.edu/catalog.php?record_id=11463.
6. Ibid.
7. National Research Council, “Research Universities and the Future of America: Ten Breakthrough Actions Vital to
Our Nation’s Prosperity and Security” (Washington, DC: The National Academies Press, 2012),
http://www.nap.edu/catalog.php?record_id=13396.
8. Peter M. Seigel, “Researchers and the Petabyte Go Global: Preparing the Next Generation of Innovators”
(presentation at the 2014 ECAR Annual Meeting, Tempe, Arizona, January 2014).
9. Anne Kelly Knowles, Dan Miller, Alex Tait, Allen Carroll, Tim Montenyohl, and Judith Nielsen, “ A Cutting-Edge
Second Look at the Battle of Gettysburg,” Smithsonian.com, June 27, 2013,